The bet whose order no person pressed
Every ordinary wager has a person at the button: the holder decides, and the holder places. A wager placed by software breaks that in a quiet way - a program takes the decision, a program sends the order, and the account it binds is still a person. This desk takes the arrangement apart: the lag between a person and a machine, the interface a program is allowed to call, the bot a customer runs, the copy product that mirrors another account, the signal that sells a selection rather than placing it, the term that requires a bet to be placed personally, how a platform notices a machine, what it does about it, and what the record shows afterwards.
- desk
- 83
- software orders sampled
- 900
- pressed by a person
- 0
- human lag
- 9.4 s
- machine lag
- 0.08 s
- flagged
- 62
A wager placed by software is an order a program sent rather than one a person pressed. Of 900 invented software orders, a person pressed the button in none; a human takes a median 9.4 seconds from deciding to an accepted price and a machine 0.08 seconds, and the difference is where the whole subject lives - because the account, the term and the payout all still carry one human name.
four stations, one name in the record - so the whole arrangement is decided by the fourth node, not the second.
the machine is not betting better; it is betting earlier, and the earlier price is the whole edge.
900 bets whose order was placed by software rather than by a person: 520 from a bot the customer ran, 240 from a copy product that mirrored another account, and 140 from a third party holding an interface key. The person pressed the button in none of them.
- software orders
- 900
- a customer bot
- 520
- a copy product
- 240
- an interface key
- 140
- pressed by a person
- 0
- channels
- 3
The time from the decision to the acceptance: a person takes a median of 9.4 seconds, a machine 0.08 seconds. The window is where the price moves, and it costs the person about 2.40 on a 100.00 stake in the sampled markets.
- human median
- 9.4 s
- machine median
- 0.08 s
- ratio
- 117.5x
- human under 1 s
- 6%
- slippage, 100.00 stake
- 2.40
- window
- the whole advantage
240 bets mirrored from another account: the copy executed at a mean 3.6% worse price than the original, diverged from the original price in 96 of 240, and paid a fee of 20.00 a month or a 20% profit share where it was offered.
- copied bets
- 240
- mean slippage
- 3.6%
- diverged from original
- 96
- divergence share
- 40.0%
- subscription
- 20.00 / month
- profit share offered
- 20%
400 signals bought by people who still place the bet themselves: 340 were placed, 60 missed, the median gap from signal to placement was 252 seconds, and only 34% were still available at the price the signal quoted.
- signals
- 400
- placed by a person
- 340
- missed
- 60
- signal to placement
- 252 s
- still at the quoted price
- 34%
- cost on 100.00
- 12.00
Every one of the 900 software orders was placed against a term that requires each bet to be placed personally: 900 of 900, in one clause the operator writes and the customer accepts at sign-up.
- software orders
- 900
- against the term
- 900
- against the term share
- 100%
- the clause
- one
- read before sign-up
- 0 of 900
- what the term requires
- personal placement
Of the 900, the platform flagged 62, revoked 48 interface keys, limited 30 accounts, closed 18 and refused 12 withdrawals - detection is the exception rather than the rule.
- flagged
- 62
- keys revoked
- 48
- accounts limited
- 30
- accounts closed
- 18
- withdrawals refused
- 12
- flagged share
- 6.9%
The 30 limited accounts were capped at a mean maximum stake of 2.00, down from 50.00; 22 of the 30 also lost an offer, and the win rate among them fell from 8.4% to 4.2%.
- accounts limited
- 30
- mean ceiling before
- 50.00
- mean ceiling after
- 2.00
- offer removed
- 22
- win rate before
- 8.4%
- win rate after
- 4.2%
All 900 orders are recorded on the ordinary internet channel with a timestamp, a price, a device and an address - and 0 of 900 name the software that placed the order or the person who wrote it.
- orders recorded
- 900
- on the ordinary channel
- 900
- naming the software
- 0
- naming the author
- 0
- a device is present
- 900
- a person is identifiable
- 0
One bet carried through: a 100.00 stake placed by a bot in 0.08 seconds at 2.05, where a person at 9.4 seconds would have taken 2.00; the price advantage was 5.00, and the account was capped at a 2.00 maximum stake afterwards.
- stake
- 100.00
- machine price
- 2.05
- human price
- 2.00
- machine return
- 205.00
- human return
- 200.00
- advantage
- 5.00
What the month did: 900 software-placed bets, 90,000.00 staked at a mean of 100.00, 62 flagged, 48 interface keys revoked, 30 accounts limited, 18 closed and 12 withdrawals refused.
- software bets
- 900
- staked
- 90,000.00
- mean stake
- 100.00
- accounts limited
- 30
- accounts closed
- 18
- keys revoked
- 48
The two components below are the whole desk at a glance: the machine chain shows the four stations between a decision and a record, only one of which is human, and the lag board puts the human path and the machine path side by side, stage by stage. Everything under them is one sample at a time.
One name, and four stations
An ordinary bet has one station: the holder decides and places. A bet placed by software keeps the holder at the top and replaces the middle. The decision may still be a person or it may be a rule; the instruction is now a program; the order is placed through the same interface a person uses; and the record names an account and a device rather than an author. The rule the desk comes back to is that every one of those stations still binds the account holder, and none of them names the software.
Why the desk stops at the placing
Four neighbouring questions belong to other desks. Who may place it for you - a friend, a paid agent or a nominee - is a person acting for another person; here the actor is a program. Custody is how the balance is held; here the balance is ordinary. The acceptance window is how the operator holds and re-quotes a price; here the price is only what the lag reaches. And the terms document is how a clause is formed and varied; here one clause is quoted and its consequence is the subject.
One order, and who is at the button
- The account is still one verified person, so every consequence lands on them.
- The term requires each bet to be placed personally, so a program is outside it by construction.
- The advantage is not cleverness, it is time: 0.08 seconds against 9.4.
- A copy product does not bet for the follower better than they could; it reproduces somebody else at a worse price.
- Detection is the exception, not the rule: 62 of 900 were flagged.